Wearable Transcutaneous CO₂ Monitor Based on Miniaturized Nondispersive Infrared Sensor

Abstract

Transcutaneous oxygen and carbon dioxide provide the status of pulmonary gas exchange and are of importance in diagnosis and management of respiratory diseases. Though significant progress has been made in oximetry, not much has been explored in developing wearable technologies for continuous monitoring of transcutaneous carbon dioxide. This research reports the development of a truly wearable sensor for continuous monitoring of transcutaneous carbon dioxide using miniaturized nondispersive infrared sensor augmented by hydrophobic membrane to address the humidity interference. The wearable transcutaneous CO₂ monitor shows well-behaved response curve to humid CO₂ with linear response to CO₂ concentration. The profile of transcutaneous CO₂ monitored by the wearable device correlates well with the end-tidal CO₂ trend in human test. The feasibility of the wearable device for passive and unobstructed tracking of transcutaneous CO₂ in free-living conditions has also been demonstrated in field test. The wearable transcutaneous CO₂ monitoring technology developed in this research can be widely used in remote assessment of pulmonary gas exchange efficiency for patients with respiratory diseases, such as COVID-19, sleep apnea, and chronic obstructive pulmonary disease (COPD).

I. Introduction

Arterial Blood Gases (ABG) carry important health information about the pulmonary gas exchange and have been commonly used in intensive care as a diagnostic tool to assess functions of cardio-respiratory system in patients. [1] The ABG test evaluates the blood oxygenation and acid-base balance by measuring the partial pressures of oxygen (PaO₂), carbon dioxide (PaCO₂), and pH in arterial blood. [2] Though being considered as the “gold standard”, arterial blood gas test is invasive, limited to central laboratory settings, and unable to provide continuous measurement. Transcutaneous blood gas monitoring is a continuous and noninvasive technique for arterial blood gas determination, [3], [4]. It has been increasingly used in oxygen therapy management, [5] fetal monitoring, [6] and clinical physiology monitoring. [7] Besides the application in the intensive care, transcutaneous blood gas analysis has been increasingly adopted in routine clinical practice, [8], [9] and used in monitoring sleep apnea and other related disorders. [10]–[13] With the rapid spread of COVID-19 across the world, the demand for measuring transcutaneous blood gas has increased due to the occurrence of arterial hypoxemia as part of this disease. [14]–[16]

Though the non-invasive transcutaneous blood gas analysis is well established, it has been mostly confined to bench-top clinical devices. A miniaturized wearable device that can be simply attached to a patient’s body for continuous monitoring of transcutaneous blood oxygen and carbon dioxide in a remote manner will revolutionize the healthcare for respiratory diseases, such as COVID-19, sleep apnea and chronic obstructive pulmonary disease (COPD). [15] In recent years, pulse oximetry has been miniaturized into wearable devices using optical sensing technologies for continuously monitoring of blood oxygen saturation (SpO₂), and is a commonly accepted method to indicate arterial blood oxygen concentration. [17]–[21] But not much has been explored to develop wearable devices for continuous monitoring of transcutaneous blood carbon dioxide.

The traditional transcutaneous CO₂ monitors (prices around a few thousand dollars) use electrochemical sensors (usually a variant of Stow-Severinghaus electrodes) to measure the transcutaneous partial pressure of CO₂ (PtcCO₂). [22], [23]. This kind of sensors determine the CO₂ partial pressure through measuring the pH of the electrolyte layer. This change in pH has a logarithmic relationship with the transcutaneous partial pressure of CO₂. But these electrochemical sensors are delicate, tend to drift over time, and require constant calibration and replacement of membrane[8] To avoid these drawbacks, various optical methods for PtcCO₂ monitoring have been reported. [24]–[26] But these optical sensors usually consist of bulky and complex components such as lamp, fiber optics, or reaction chamber, making them not compatible with miniaturized wearable platform. Moreover, traditional transcutaneous CO₂ monitors heat up the area of skin where the sensor patches are placed to improve the visibility and diffusion of CO₂ in the contact area. This procedure is known to be painful since it damages the skin and the associated tissue underneath in long-term monitoring. Thus, a periodic change of the sensor patch location is required in clinical use to avoid skin burning[27] To mitigate this risk, studies have shown that using the diffusion rate of CO₂, rather than the equilibrating concentration of CO₂, as an indicator to determine the partial pressure of transcutaneous CO₂ can avoid skin heating in PtcCO₂ monitoring. [28], [29] Though promising results such as trends in CO₂ concentration with varying levels of activity of the subject have been demonstrated, bench-top bulky NDIR sensors are still adopted in these systems. These systems also require using pure N₂ to periodically purge the dead space in the sensing chamber, making them not compatible with the wearable sensing platform.

In this research, we have developed a miniaturized nondispersive infrared (NDIR) sensor-based wearable wristband device for continuously and noninvasively monitoring of transcutaneous blood CO₂. This miniaturized NDIR sensor is featured by its accuracy, longevity, robustness, and low-power consumption. We have also developed a hydrophobic membrane with high CO₂ permeability to mitigate the humidity interference on the performance of the NDIR CO₂ sensor. By using the diffusion rate-based detection principle for signal processing, the wristband achieves reliable and continuous transcutaneous blood CO₂ tracking without skin heating. The performance of the NDIR CO₂ sensor-based wristband has been validated against end-tidal CO₂ monitor by simultaneously monitoring the arterial blood CO₂ change of a subject over various activities throughout the day.

II. Materials and Methods

A. Integrating Miniaturized NDIR Sensor to Wristband

Nondispersive infrared (NDIR) sensor is the most common kind of CO₂ sensor. It detects CO₂ by measuring the IR absorption fingerprint of CO₂ molecules at the wavelength of 4.26 microns and using Beer-Lambert’s law to quantify the CO₂ concentration in the gas sample. Tremendous efforts have been spent to miniaturize the size of the NDIR CO₂ sensors in the past decades. With the advances of optics and waveguide design, longer optical light path has been dramatically folded, allowing the production of miniature NDIR CO₂ sensors with higher sensitivity and reliability. We have used a Cozir NDIR CO₂ sensor to build the wearable transcutaneous CO₂ wristband (Fig. 1(c)). The Cozir NDIR CO₂ sensor is based on solid-state LED technology and featured coin-sized form factor (ø20.9mm × 18.1mm), light weight (4g), low power consumption (3.35mW @ 3.3V), fast response time (30s), high accuracy (±70ppm), wide dynamic range (0–5%), and long lifetime (>15 years). All these features make it very suitable for wearable sensing platform. This sensor has been used for healthcare, safety, and aerospace etc. The sensor has a built-in auto-calibration system, which uses the default fresh air CO₂ concentration of 400ppm to calibrate the sensor. The sensor can constantly send streaming measurement data twice per second via its UART interface. Because of the high accuracy and real-time sensing capability, the Cozir NDIR CO₂ sensor can be directly integrated into an enclosed chamber on the wristband for continuous transcutaneous CO₂ monitoring without implementing any active gas sampling components to the system, as shown in Fig. 1.

Fig. 1.

Wearable Transcutaneous CO₂ Monitor Based on Miniaturized Nondispersive Infrared Sensor. (A) Image of the integrated wristband. (B) The zoom-in shows cross-section view of the gas chamber, sealing O-ring, and wristband body. (C) Miniaturized Cozir NDIR CO₂ sensor.

The integrated wristband for continuous transcutaneous CO₂ monitoring is illustrated in Fig. 1. The wristband consists of a 3D printed plastic body that accommodates the Cozir NDIR CO₂ sensor. The CO₂ sensor is exposed to a miniaturized gas chamber (volume of 1.0 ml), where the transcutaneous CO₂ released from the skin can diffuse directly in and accumulate in a fixed space. An O-ring placed over the gas chamber acts as a cushion between the watch and the skin, which also provides air-tight sealing to avoid any gas leakage from or into the ambient air.

The Cozir NDIR CO₂ sensor was mounted on a circuit board that hosted UART connectivity to report the sensor data. The sensor was connected to a Windows PC using UART-USB connector to power the sensor circuit. The “GasLab” software was used to collect, view and save the sensor data. The software also provided an interface to calibrate the sensor when necessary.

B. Using Hydrophobic Membrane to Mitigate the Humidity Interference

Transcutaneous blood gas is humid, because water molecules leave human skin through diffusion and sweating. [30], [31] Water molecules interfere non-dispersive infrared gas sensors [32] since the absorption spectrum of water shows high extinction coefficient at wavelengths close to 4.26 microns. [33] The initial tests were performed by strapping the watch to a subject’s wrist. The subject was in resting condition and seated in a room with an ambient temperature of 25 °C. The wristband was worn on the same wrist for all the tests. As seen in Fig. 2, the trend in the buildup of CO₂ within the O-ring were not reproducible due to the interference of humidity and condensation from the skin.

Fig. 2.

Response of Cozir NDIR CO₂ sensor to transcutaneous CO₂. (a) Measurements with interference from humidity and condensation. (b) The temperature and humidity profile of the setup when the wristband was strapped to the skin.

A hydrophobic membrane could be effective to mitigate the humidity interference for NDIR CO₂ sensor for transcutaneous blood gas measurement. Hydrophobic membranes are polymers that have low permeability to water vapor. They have been widely used in electrochemical sensors [34] and colorimetric sensors [35] to repel the water molecules. But when using them for transcutaneous gas monitoring, low permeability to water, high permeability to CO₂, and hydrophobicity (contact angle to water droplets) are the important properties that need to be considered. Membranes like Polytetrafluoroethylene (Teflon) and Polydimethylsiloxane (PDMS) have better permeability to CO₂, as shown in Table I. Traditional transcutaneous gas monitors use Teflon as the hydrophobic membrane with the associated CO₂ sensors. [36], [37] Compared to PDMS, Teflon has low permeability coefficients to both H₂O and CO₂. Though highly effective in reducing the humidity effect, Teflon also decreases the sensitivity of the sensor for transcutaneous CO₂ monitoring. Fig. 3 shows the response of NDIR CO₂ sensor to transcutaneous CO₂ measured on the arm of a subject with Teflon or the PDMS membrane. It is clear that the response from the sensor with Teflon is only 1/3 the response from sensor with the PDMS membrane. Compared with the NDIR CO₂ sensor response in Fig. 2, it is evident that the sensors with hydrophobic membranes show well-behaved response curves, which reflect the mass transportation process of CO₂ molecules through the membrane. The sensing signal increases quickly at the beginning as the CO₂ molecules diffuse from the gas chamber to the sensor side due to the concentration gradient and then reach a plateau when the concentration equilibrium is established. PDMS is used in applications that require high permeability of gases, [38] which is important for fast-response sensing. Since real-time transcutaneous CO₂ monitoring requires high sensitivity and fast response time, we have chosen PDMS as the hydrophobic membrane in our wristband design.

TABLE I.

H₂O and CO₂ Permeability Coefficients Comparison Between Teflon and PDMS Membranes. [39]–[42]

Membrane			Permeability Coefficient (Barrer*)
Name	Structure & Appearence	Contact Angle (°)	H2O	CO2
Polytetrafluoroethylene (Teflon)		115.3	425	520
Polydimethylsiloxane (PDMS)		113.5	45,000	4000

^*Barrer = 10⁻¹⁰ cm³ . cm/cm² . s . cmHg

Fig. 3.

Evaluation of hydrophobic membranes for NDIR CO₂ sensing in the wristband. Response of the CO₂ sensor with the Teflon and PDMS membranes when tested transcutaneous CO₂ on the arm of a subject.

The membrane was synthesized by using Sylgard^™ 184 Silicone Elastomer Kit from Dow Corning Corporation. A petri dish with the diameter of 50mm was used and 0.9 g elastomer solution was spread across the base of the petri dish. The solution was cured at 45°C overnight to produce a membrane with a thickness of 0.15mm[43] The Teflon (Polytetrafluoroethylene) membrane (thickness 0.1 mm, size 100 × 100 mm) was purchased from McMaster-Carr.

C. Bench Tests

Bench tests were performed to understand the influence of the membrane on transcutaneous gases as well as the response behavior of the NDIR sensor. The offline bench test setup, as illustrated in Fig. 4, allowed the evaluation of PDMS membrane over various concentrations of CO₂ under different conditions. The diffusion chamber on the top provided an inlet and outlet for the gas being actively pumped at a flow rate of 0.8 L min-1. A preparation phase of 5 minutes was used to allow buildup of CO₂ inside the diffusion chamber and eventually diffused towards the CO₂ sensor. The PDMS membrane was fixed on the other end of diffusion chamber. The membrane and CO₂ sensor were separated by a gap with the volume of ~1.0 ml. The chamber was purged with pure N₂ to remove any traces or residue of contaminants that remained behind between tests. The test setup was maintained at a relatively constant room temperature of 25 °C. A set of 6 different concentrations of CO₂ in the range of 0.75% to 4% were generated using a mixture of 4% CO₂ and N₂ from gas cylinders by Praxair and Matheson, respectively. The Cozir sensor was calibrated with 0% and 4% CO₂ before inserting it into the bench setup. To test the sensor response to humidity, the gases were bubbled through a bench top water bath at room temperature of 25°C, which generated humid sample gases with 85% RH. All gas samples with different CO₂ concentrations were tested back-to-back to ensure consistency between results.

Fig. 4.

Schematic of the setup for bench tests.

D. Testing the Transcutaneous CO₂ Wristband Against End-Tidal CO₂

The subject study was approved by the Institutional Review Board of Arizona State University (IRB reference protocols # STUDY00013474). The wristwatch was evaluated against End-tidal CO₂ (EtCO2), since it was a good representative of arterial CO₂. [44]–[46] EtCO₂ was measured using breath-based CO₂ monitor by VacuMed (carbon dioxide analyzer, model 17630), which was a fast-response NDIR based sensor using Nafion tubing to mitigate the humidity interference. The inlet of the CO₂ monitor was connected to a pneumotach based mask setup, which was worn by the subject for normal breathing under resting conditions. The analyzer was connected to a data acquisition system as its data logger. Then, the subject strapped on the wristband device and connected the UART-USB interface. The data loggers simultaneously collected sensor data from both the CO₂ monitor and the wristband device. The subject was asked to sit down, rest, and breathe normally during the test until the wristband reading reached a plateau. From this point, the sensor data were used to analyze the performance of wristband device.

III. Results and Discussion

A. Analytical Performance of the Transcutaneous CO₂ Wristband

As seen in Fig. 5 (a), the CO₂ response curves from the PDMS membrane covered NDIR sensor showed sharp increase first and then reach plateaus. The plateau readings were proportional to the concentrations of CO₂ diffusing into the NDIR sensor, as shown in Fig. 5 (b). The sensor response to humid CO₂ gas sample generated from water bath setup was offset from the response to dry gas sample by a linear scaling factor of 0.797, shown in Fig. 5 (c).

Fig. 5.

Performance of Cozir CO₂ sensor with PDMS membrane. (a) Response of the sensor for varying conc. of CO₂ pumped into the diffusion chamber. (b) Correlation between the plateau and the conc. of CO₂ shown in (a). (c) Correlation between the response of Cozir sensor to dry and humid CO₂. (d) Rise in CO₂ across the PDMS membrane after purging the sensor with N₂. (e) Linear fit of the CO₂ trends shown in (d). (f) Correlation between the slope of the linear fit and absolute concentration of diffusing CO₂.

It was observed that the slopes at the rising stage of response curves also showed a linear correlation with the CO₂ concentration, as seen in Fig. 5 (d)–(f). This indicated that it was not necessary to wait for the sensor to reach a plateau to determine the CO₂ concentration, which could be used to reduce the response time. Considering that active sample delivery components, e.g., pump required extra power consumption and could complicate mass transportation of transcutaneous CO₂ in the wearable device, we used the plateau response from the NDIR sensor to track the transcutaneous CO₂ when testing the sensor on subjects.

B. Validation of Transcutaneous CO₂ Wristband With End-Tidal CO2 With Human Subjects

End-tidal CO₂ (EtCO₂) is a reliable reference technique to validate the NDIR sensor-based wearable wristband device, since partial pressure of end-tidal CO₂ shows a strong positive correlation with arterial and transcutaneous CO₂ (TcCO₂). [47], [48] The end-tidal CO₂ was measured using fast-response breath analyzer (shown in Fig. 6 (a)) connected to a pneumotach based face mask and a sampling tube. The end-tidal CO₂ and wristband CO₂ signals were simultaneously monitored in real time for point-to-point comparison. As shown in Fig. 6 (b) & (c), the EtCO₂ was computed by selecting the concentration of CO₂ at the end of exhalation cycle right before the flip towards inhalation began. For transcutaneous CO₂ monitoring, the sensor was placed on the subject’s wrist using watch straps to tightly hold on to the skin to avoid leakage. After the sensor in the wristband reaches a plateau, the EtCO₂ and TcCO₂ were monitored for another 10 minutes to ensure the subject and sensor responses were stable. The sensor signals were periodically averaged over a window of 2 minutes to remove random noise and retrieve an overall trend.

Fig. 6.

Validate the transcutaneous CO₂ wristband with end-tidal CO₂ on human subject. (a) VacuMed carbon dioxide analyzer, model 17630, is used to track end-tidal CO₂ in real time. (b) The real time profile of exhalation carbon dioxide by the subject. (c) Zoom-In plot of a single respiratory cycle with an indication of end tidal CO₂. (d) Profiles of transcutaneous and end-tidal CO₂ from a subject under various activities (i.e., baseline/plateau, consumption of black coffee and cookie, and hyperventilation).

The validation test started with the subject at resting (fasting for 12 hours). The NDIR sensor reached a plateau with stable reading at resting condition. Then, the subject consumed black coffee, which caused both the end-tidal CO₂ and transcutaneous CO₂ readings to increase. The CO₂ readings peaked and then started to decrease. There was a lag time of ~5min between the transcutaneous and end-tidal CO₂, as seen in Fig. 6 (d). Between 45 and 55 min, a hyperventilation was performed by the subject. During the hyperventilation, both transcutaneous and end-tidal CO₂ quickly decreased. After the hyperventilation both transcutaneous and end-tidal CO₂ gradually increased and then reached stable readings. Following the activity change during the test, the profile of transcutaneous CO₂ corelated very well with the trend of end-tidal CO₂.

C. Use Wristband to Track Transcutaneous CO₂ During Daily Activities in Free-Living Conditions

The wristband sensor was also evaluated in a field test to track the transcutaneous CO₂ profile of a subject during daily activities in free living conditions for about 4.5 hours. Before the field test, NDIR sensor-based wearable wristband device was first calibrated against the end-tidal CO₂ using the VacuMed carbon dioxide analyzer. As suggested by the calibration curve in Fig. 7 (a), the CO₂ readings from the wristband device showed highly linear correlation with the end-tidal CO₂ readings. By using this calibration plot, the signal from the wristband device can be converted into transcutaneous CO₂ readings in partial pressure (mmHg). As shown in Fig. 7 (b), the transcutaneous CO₂ changed in accordance with the daily activities from the subject. After consuming rice, beans, and cheese for lunch, the transcutaneous CO₂ of the subject showed a quick rise. This is expected since consumption of food increases the metabolism, thus increases CO₂ release. Higher levels of CO₂ were reported after lunch, which is consistent with the trends reported in other previous work. [28] Further consumption of black coffee caused an expected rapid increase in CO₂ as well. The result of the field test clearly demonstrated the feasibility of this NDIR sensor-based wearable wristband device for passive and unobstructive tracking of transcutaneous CO₂ in free-living conditions.

Fig. 7.

(a) Calibration of transcutaneous CO₂ wristband with end-tidal CO₂. (b) Tracking the transcutaneous CO₂ profile of a subject during daily activities in free-living conditions with the wristband device.

IV. Conclusion

In this research, we have developed a miniaturized nondispersive infrared (NDIR) sensor-based wearable wristband device for continuous transcutaneous CO₂ monitoring. To address the humidity interference in transcutaneous CO₂ detection, a PDMS membrane has been synthesized and integrated to the NDIR CO₂ sensor. The PDMS membrane assisted NDIR CO₂ sensor shows well-behaved response curve to humid CO₂ with its plateau reading linearly corelated with the CO₂ concentration. The NDIR sensor-based wearable wristband device has been validated against end-tidal CO₂ in human test. The profile of transcutaneous CO₂ correlated very well with the expected activities-specific end-tidal CO₂ changes over the test. The feasibility of the NDIR sensor-based wearable wristband device for passive and unobstructive tracking of transcutaneous CO₂ in free-living conditions has also been demonstrated in the field test. This research has provided a simple, cost-effective, and heating-free approach for tracking of transcutaneous CO₂ on a wearable platform. The miniaturized wristband device demonstrated in this research can be simply attached to the arm of patient with respiratory diseases, such as COVID-19, sleep apnea, and chronic obstructive pulmonary disease (COPD) for remote pulmonary gas exchange efficiency monitoring. To further enhance the long-term stability and the response time of the NDIR CO₂ sensor-based wearable wristband device, technical innovations in self-calibration, diffusion optimization, signal processing, and algorithms improvement need to be explored and evaluated. Through the integration of Bluetooth chip and battery to the sensing system, a stand-alone transcutaneous wristband can be designed for both clinical use and home use. Furthermore, the wearable transcutaneous CO₂ monitoring technology developed in this research can also be easily integrated with miniaturized and off-the-shelf pulse oximetry sensors for comprehensive assessment of arterial blood gases in routinely clinical practices.